Hybrid memory infra for agents at scale

When an agent finishes a task, Memorable stores how it was done — the files that changed, the commands that verified it, real outcomes — and replays that knowledge when a similar task returns. Extraction is deterministic (no LLM anywhere), storage is on the user's side, and consent is fail-closed. Measured: 19% fewer agent turns, replicated across two independent n=25 runs (p < 0.0001), the second receipt-bound.

CLI

Command reference

init, consent, ingest, recall, the viewer, doctor — every command with what it does and what it refuses to do.

API

Extraction API

Self-serve keys, one deterministic endpoint, request ids on every response. Integrate any harness directly.

Integrations

gbrain · QM

Fully automatic on both: session capture, opt-in relays, minimized egress, kill switches.

Agents

Let your agent drive

One copyable prompt teaches any coding agent the whole loop.

Quickstart

Two minutes, no account. The default backend is a standalone local store; on a machine running gbrain, use memorable init gbrain instead for automatic session capture.

install + opt in

npm install -g memorable-cli
memorable init      # auto-issues an API key, picks the local backend
memorable enable    # explicit write consent — nothing is stored before this

store a procedure, then find it

memorable ingest trace.json
memorable recall "rotate the TLS cert"
memorable show <slug>     # guarded, injection-safe rendering
memorable graph           # browse everything at localhost

For agents

Telling a coding agent "use memorable" is enough — every step is a plain CLI call it can run itself. Make it durable per project:

one line

memorable agents-md >> AGENTS.md

Both the CLI and API pages open with a copyable agent prompt; the assistant in the corner answers questions from these docs without sending anything anywhere.